TopoLabs
[About]

Making inference a commodity.

We are building an open market for AI compute. Anyone with a GPU can serve a model, anyone with a wallet can buy tokens, and every request settles in seconds.

[Our story]

Compute is everywhere. Access is not.

There are millions of capable GPUs in the world, and many of them sit idle for hours a day. Meanwhile, teams building with AI juggle five vendor contracts, rate limits and invoices just to call the models they need.

Topo Labs started in 2024 with a simple bet: if serving a model were as easy as running a node, and paying for a token as easy as sending one, inference would become a commodity. Cheaper, more resilient and open to anyone.

Today Topo routes every request to the best available node, verifies the result and settles the payment per token in USDT, ETH or SOL. One API, no sales call, no account manager required.

GPU nodes online
2,340
Regions served
41
Models on the network
120+
Tokens served to date
1.2T
[Principles]

How we build the network.

Four rules we hold ourselves to when we design the protocol, the product and the company.

  • 01

    Open by default

    The router, the node client and the verification spec are open source. Anyone can read exactly how a request gets served.

  • 02

    Verifiable

    Every response carries a receipt: which node served it, what it cost and an attestation you can check on-chain.

  • 03

    Priced per token

    No seats, no minimums, no annual contracts. You pay for the tokens you use and nothing else, settled in seconds.

  • 04

    Operator first

    Nodes are the network. We publish payouts, keep fees flat and ship the tooling operators actually ask for.

[Team]

A small team with a large surface area.

Engineers and operators from cloud infrastructure, crypto protocols and ML research.

  • Ana Ribeiro

    Lisbon

    Co-founder, CEO

    Ran GPU capacity planning at a hyperscaler. Thinks compute should trade like bandwidth.

  • Daniel Okafor

    London

    Co-founder, CTO

    Wrote consensus code for two L1s. Leads the router and the settlement layer.

  • Mei Tanaka

    Tokyo

    Head of Network

    Onboards node operators and owns uptime across all 41 regions.

  • Lukas Brandt

    Berlin

    Staff Engineer, Inference

    Makes open-weight models run fast on hardware nobody else bothers to support.

  • Priya Raman

    Bengaluru

    Head of Developer Experience

    Owns the SDKs, the docs and the five-minute quickstart. Reads every support thread.

  • Sam Whitfield

    Toronto

    Protocol Research

    Designs the verification scheme and the staking economics that keep nodes honest.

Remote-first. 14 people across 9 countries. Async by default, two overlap hours a day.

Backed by
  • Northwind Ventures
  • Basalt Capital
  • Fieldline
  • Orbit Fund
[Contact]

Build the network with us.

Questions about the protocol, partnerships or running nodes at scale? We read every message.

Talk to usVisit careers